What kind of git commits does this skill create?

The skill performs atomic commits for every successfully completed task. It follows the Conventional Commits standard (e.g., feat, fix, docs) and includes the plan name and task name in the commit message.

How does Long-Run Implement prevent AI performance degradation?

It utilizes a 'fresh context' architecture, spawning a new subagent for each individual plan. This ensures every task starts with 0% context usage and peak token availability, avoiding the quality drop-off seen in long conversations.

Can I resume a task if the process is interrupted?

Yes, the skill maintains full state persistence in a local .long-run/ directory. If execution is interrupted, simply run the skill again on the same spec folder and it will offer to resume from the last saved checkpoint.

Can I review the generated plan before implementation begins?

Absolutely. By using the --prepare-only flag, the skill will stop after Phase 2 (Plan Preparation), allowing you to inspect and modify the individual plans before any code is actually written.

Long-Run Implement

Name: Long-Run Implement
Author: fiatkongen

byfiatkongen

0•

Productivity & Workflow

Executes complex implementation plans autonomously through fresh context subagents and atomic git commits.

The Long-Run Implement skill enables Claude Code to handle large-scale implementation tasks by orchestrating a series of smaller, high-quality execution phases. It addresses the common issue of context window degradation by spawning a fresh subagent for each plan component, ensuring peak performance and accuracy across long-running projects. With built-in state persistence, automatic deviation handling, and per-task atomic commits, it provides a robust framework for developers to delegate multi-step coding plans while maintaining high quality and easy recovery from interruptions.

Key Features

010 GitHub stars

02Orchestration Integration: Leverages orchestration.yml to assign specialized agents (e.g., backend vs. frontend) to specific tasks.

03Per-Task Atomic Commits: Automatically creates git commits for every completed task using conventional commit formatting.

04Fresh Context Spawning: Starts each plan with a new subagent to maintain peak coding quality and 200k token availability.

05Full State Persistence: Saves execution progress to disk in a .long-run directory for seamless resumption after interruptions.

06Automatic Deviation Handling: Resolves minor implementation issues autonomously while logging architectural blockers for user review.

Use Cases

01Implementing multi-file feature specifications from a detailed architectural roadmap.

02Performing large-scale codebase refactoring that requires sequential, verified steps.

03Executing autonomous system migrations with persistent progress tracking and minimal checkpoints.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add fiatkongen/saurun-marketplace long-run-implement

For use in Claude.ai and ChatGPT

Download Skill